Physical Realization of Pseudorandom Perturbations for Human System Identification
Author(s)
Qiu, Yingxin
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Understanding the neuromechanics of human movements has a significant impact on our health, work, and recreation. An effective approach to model the dynamics of human movements is by performing human system identification experiments. Motion platforms are often utilized in such experiments to provide physical perturbations. To ensure the performance of human system identification, the motion control system must be integrated to be able to generate high-quality physical perturbations.
Ideal properties of perturbation signals as well as how to generate reference commands to the motion control platform have been extensively studied in the field of system identification. When it comes to their physical realization, trajectories as the outputs of the motion control platform must be carefully produced to satisfy the specifications. However, there is a void in system engineering research to ensure the quality of produced physical perturbations with respect to their frequency-domain characteristics. Since perfect physical realization of perturbation signals is virtually impossible due to the dynamics of motion control hardware, the motion control system must be configured to account for the perturbation performance degradation issues.
This research presents methods to generate two types of physical perturbations with enhanced spectral properties. The first is the pseudorandom sequence, which is used when prior knowledge regarding the target human system is not available. Novel metrics were proposed to evaluate the perturbation quality as well as the performance of the motion control system used for physical realization. The second type is the optimal multisine perturbations, which requires prior knowledge of the target system. The optimization problem of the optimal multisine design was reformulated to take the device dynamics into consideration. In practice, pseudorandom perturbations can be applied first to acquire initial guesses, followed by the design and application of optimal multisine perturbations to refine the existing models. For both perturbations, human experiments were conducted to verify that the proposed methods improved the spectral properties of the physical perturbations and, as a result, improved the modeling accuracy.
Sponsor
Date
2023-01-10
Extent
Resource Type
Text
Resource Subtype
Dissertation